Skip to content

Decoupling social media

Updated: at 03:22 PM


The damage social media platforms inflict on our society is now well known. In this post, I describe how we can change the social media architecture to improve experiences and outcomes. To analyze social media services, let’s look at five common functions they all implement and see how we can improve them. When we take a closer look at social media, we see that they:

Let’s look closely at each function’s problems and see how we can address them with a decoupled social media platform.


In today’s social media, you sign up for a platform and pick an identity, typically called a ‘nickname’ or ‘handle.’ With some services, you establish relationships with family, friends, or colleagues. With others, you discover people or businesses on the platform, and you start following or subscribing to them. If you are a creator or influencer on a social media platform, you build up an audience of followers, and the platform rewards you for your content.

If you switch to a different platform, you may be able to use the same ‘nickname’ or ‘handle,’ but no one knows whether you own that identity on both platforms or whether one of the two accounts impersonates you. Additionally, when you switch to a different platform, you must recreate your network containing all the people who follow you or who you follow, assuming these people are on the new platform in the first place. If you are a creator, this is a significant problem, as your livelihood may depend on your audience.

The lack of seamless migration of identity and network makes switching platforms hard and strengthens the ‘network effect’; the more people use a platform, the more valuable the platform becomes to those people. social media identity

Of course, there are technical solutions to this problem, but social networks do not want to facilitate seamless migration to different platforms. Because of the benefits of the network effect, they know that any competitors delivering similar experiences and features are all but doomed to fail.

If we start looking for solutions to these problems, we see that decoupling various functions from social media services can create better outcomes. It will also reduce the impact of the network effect and thus foster more competition between services.

For identity, we should be able to prove that we own our identities across different services. Additionally, a third party can provide a service that verifies that we are who we say we are and provide us with proof that this is true. We can then share this proof on social media services so that people who trust that third-party identity verification provider know who they are dealing with. With such features, we can start following a creator across all social media services where they publish content. That will allow both us and the creators to migrate from one service to another more easily.

Content discovery

Social media run recommender algorithms that put content in your feed and search results based on your previous content consumption. Platforms design these algorithms to maximize your engagement with their platform. The more engaged you are, the longer you stay online and the more ads they can show you. Research has shown that people remain on a site or in an app longer when presented with content that enrages or scares them. Hence, the algorithms prefer such content, negatively affecting people’s mental health and society. content discovery

Of course, social media can change these algorithms or give you a choice of algorithms, but that would affect their ad revenue, so it will not happen.

We need to create an ecosystem of open-source recommender software and create social media platforms that enable us to select our recommender algorithm to match the content we want to consume.

Content moderation

Moderation on social media has become a hugely controversial subject. The first issue with content moderation is that social media platforms define their policies to moderate content, and people increasingly select their platforms based on these policies. The resulting fragmentation of audiences means that people are even more at risk of ending up in an echo chamber that does not expose them to well-reasoned opinions of others with different views. Second, as we have seen with Twitter, owners of social media can change content moderation policies, upsetting their audience. The audience might want to migrate to different platforms, but the network effect impairs this migration. Third, various actors abuse content moderation policies. There are many examples here, with foreign governments trying to affect elections, people spreading misinformation, and creators reporting that platforms such as Instagram and YouTube block their content because it gets flagged by people for bogus reasons.

Finally, creators self-censor content to minimize the chance of their content getting blocked, either by content moderation algorithms or by getting flagged by people. Creators are concerned that not just a tweet, podcast, or video but their whole account might get blocked, causing them to lose all their income on the platform where they have established their following. This self-moderation impedes the exchange of ideas and opinions on the platform.

Social media believes it can solve this problem with (artificially) intelligent algorithms and customer support teams. But we repeatedly see that humans can still outsmart algorithms, and scaling up support teams costs a lot of money, so platforms continue with poor content moderation decisions.

Here again, we can decouple the process from the social media platform. Different people, NGOs, or companies can provide content moderation services. They would publish their moderation policies, and we can select the organizations with the policies that match our needs. Sometimes, we may have to pay for this service as a subscription, or we are asked to donate. In other use cases, the content producer could pay proactively for the moderation of their content so that recommender algorithms know it is safe to add that content to the feed.

Content monetization

Social media has given rise to the era of creators and influencers. These creators grow large audiences on a platform and either directly or indirectly benefit from monetizing their published content. The first problem with this is that it makes creators, for a large part, dependent on these platforms. They do not influence the decisions of these companies. If they disagree with a decision, they typically do not have the option to switch to a different platform, as they would have to rebuild their audience from scratch.

The second problem is that these platforms take a large share of the revenue. YouTube retains a 45% cut of all ad revenue for a video. Twitch unilaterally decided last year to change the revenue split with some of their biggest streamers, taking 50% of revenue instead of 30%.

Today, creators submit content to a platform, and the platform then monetizes that content and may give the creator a share of the revenue for that content. Content monetization strategies include ads, subscriptions, and micro-payments.

When we decouple ad delivery, the creator has a direct relationship with an advertiser or their ad agency. In this model, the creator and the advertiser can negotiate directly about rates, and the creator can choose the advertiser:

Advertisers can select technologies or agencies:

There is the risk of fraud in this architecture, just like there is in the current monetization architectures. However, in the decoupled architecture, creators and advertisers have direct control over who their counter party is, and they can pick them based on who is best at reducing fraud. This is an improvement over the current models where they have to depend on the social media platform without any control or visibility on how to prevent ad fraud.

Social media uses two models for subscriptions: Twitch uses a model where you subscribe to a specific creator, while the YouTube Premium subscription gives you ad-free access to content from all creators on their platform. The decoupling model for the Twitch use case is straightforward; the relationship is directly between the creator and their followers, and the creator manages access to their content for the subscribers. The implementation of the second subscription model is more complex. People would sign up with a third-party subscription processor for a fixed amount per month. When they consume content, their web app submits a digitally signed confirmation to the creator that they have viewed their content. At the end of each billing period, each creator presents the play-out records they have received to the micro-payment processor. The payment processor then allocates payouts from the subscription fees to each creator based on the play-out records the subscribers have signed.

Decoupling for micro-payments is straightforward: There would be multiple processors. The creator decides which processors they want to support, and people wanting access to the content can pick one of the processors that the creator has chosen to work with.

Privacy and hosting of your data

Social media platforms collect vast amounts of data about you and use that not just to recommend content for you but to allow advertisers to target ads to you. In a deposition for a lawsuit, Facebook’s engineers admitted that there is no comprehensive documentation about what data Facebook collects and where it is stored in the Facebook infrastructure. The feature to download your data on their website does not include all the data about you. If Facebook does not know where your data is, how can it protect it from falling into the wrong hands?

The solution here is to avoid storing your data in the infrastructure of social media services in the first place. Byoda is a ‘personal data server’ that I’m developing. With Byoda, you store all your data and content in your Byoda ‘pod.’ Your pod would typically run in the cloud, although installing the pod in your home network is also possible. Byoda introduces an innovation other personal data servers do not have: The ‘data contract.’ This data contract specifies all the data that will be stored for the service in your pod, together with a specification of who gets access to this data. So, services only get to store data in your pod if that data complies with the data contract. The pod stores the data and makes it available to others based on the access rights described in the data contract. In this model, you know what data about you is collected because the contract describes it. You can change or delete any information anytime, as you have complete control over your data. In the Byoda model, while a service typically has access to some of the data in your pod, the service is not permitted to store your data long-term in their data centers. They are only allowed to cache the data for a short period, and after the cache expires, their systems have to wipe the content automatically. They then have to fetch the information from your pod again. This caching requirement means that when you change or delete some of your data, you know that the old data will automatically disappear from the network after the cache period expires.

The pod also supports P2P data access; for example, suppose you want to throw a big party, then you could send an invitation message to your pod to be distributed to all the friends of the friends of your friends. Your pod will then send the message to the pods of your friends, and those pods will then forward it to the pods of the next layer of friends, and so on.

When social media services are no longer the middlemen in content monetization, they will not pay for the cost of distributing content. The Byoda pod can host text, images, audio, and video content and make them available to others. In many cases, hosting the content on your pod is a great solution. However, creators who expect their content to go viral need to consider the bandwidth costs of cloud providers, and they may want to use a lower-cost CDN to deliver the content instead. Finally, the pod will, in the future, support P2P file-sharing to enable low-cost content distribution. P2P distribution can be the right solution if the high performance of a CDN for instant content consumption is not needed. What remains of social media services?

This decoupling of functions does not mean social media services as we know them today would become obsolete. They still would have to develop the web app. They can provide centralized APIs for search and discovery. They can also offer content management functions, such as YouTube Studio, where creators upload and prepare their videos for distribution. Social media services also perform the branding and marketing of the service, and they can partner with creators to help them grow their audience. The service provider can choose to decouple only some of the functions, and even for those decoupled functions, the service can make its implementation available in competition with third parties.

Additionally, while P2P data networking is an exciting architecture, there are functionalities where a centralized capability makes a lot of sense. P2P networking is typically not known for its low latency. Centralized functions can provide low latency where the same function implemented as P2P incurs higher latencies that impair the user experience. Some examples of tasks that benefit from centralization are searching and discovering content not hosted by any pod in your (extended) social network and aggregating and anonymizing usage data from the pods to make them available to recommender algorithms.


In summary, the decoupled architecture provides the following benefits:

The decoupled architecture that I advocate for will be challenging to establish. Not only do we need new or existing social media services to adopt the architecture, but we will also need organizations for identity verification and content moderation. The open-source community will have to develop recommender algorithms. Advertisers and their agencies must implement systems to work directly with creators to deliver ads. Payment processors need to support subscription and micro-payment services. Cloud providers must make running your pod seamless, easy, and cheap.

And are people waiting for this architecture? Influencers, creators, advertisers, and payment processors benefit from the monetization features but need audiences on social media services that implement the decoupled architecture. Most of the current social media services audience is interested in something other than what is happening behind the scenes and are happy with the current social media experience. But there is a group of people who worry about content moderation, and the size of that group is increasing. Many have tried using different platforms, but the network effect of the big social media platforms makes switching to alternative platforms painful. At the end of the day, most people will use the platform where the content is. When creators make content available on the decoupled architecture, people will follow. To drive adoption, we need to persuade influencers and creators using the monetization capabilities to make their content available on the new architecture, as that will enable their audiences to migrate.